Dietary Selenium and β-Carotene Contribute to the Inverse Association Between Antioxidant Mixtures and Stroke: Evidence From NHANES Using Mixture Models and Machine Learning

Traditional single-factor analyses often fail to comprehensively evaluate the synergistic effects among dietary nutrients. This study aimed to examine the association between combined dietary intake of six antioxidant-related nutrients and the odds of stroke. Using the NHANES database (n = 37,746), we applied 1:1 propensity score matching (PSM) to adjust for baseline confounders, establishing a matched sub-cohort of 2750 individuals. Targeting six antioxidant-related nutrients (vitamins A, C, and E, β-carotene, selenium, and zinc), we used weighted quantile sum (WQS) regression, quantile g-computation (qgcomp), Bayesian kernel machine regression (BKMR), and four machine learning algorithms to quantify joint exposure effects and estimate the relative contributions of each component. Across these complementary analytical frameworks, higher combined intake of the six nutrients was consistently associated with lower odds of stroke. In the WQS and qgcomp models, selenium and β-carotene received the largest weights in the observed inverse mixture association. BKMR assigned the highest posterior inclusion probabilities (PIPs) to selenium (0.988) and β-carotene (0.975), and these two nutrients also ranked as the top two features across all four machine learning algorithms. Non-linear analyses suggested curvilinear inverse associations for selenium and β-carotene, with attenuation of the inverse association at higher intake levels. Furthermore, single-variable effects were parallel across background exposure quantiles, with no statistical evidence of departure from additivity among the six nutrients. In conclusion, higher combined intake of this nutrient mixture was associated with lower odds of stroke. These findings support further investigation of antioxidant nutrients as a dietary mixture rather than as isolated components. Before preventive recommendations can be made, prospective cohort studies with repeated dietary assessment are needed for stroke prevention.